The Armbot is an amazing arm exoskeleton
which I developed in a series of
prototypes
from 2012 to 2017.
The movie shows the "final", all-singing, all-dancing
prototype #12.
The Armbot is a remarkably effective, low cost and light­weight device, thanks to a few simple but unusual design ideas. This website documents some of its more interesting design features.
The Armbot did not go into production for company political reasons. Since every cloud has a silver lining, the patenting process was also abandoned. The design ideas presented here can therefore be freely copied, and I would encourage anyone to do so.
Armbot final prototype ( 2017 )
The single most important idea in the Armbot is that it drives the human joints in torque only. This includes elbow flexion, as well as the shoulder degrees of freedom : shoulder abduction, shoulder flexion, and internal-external rotation.
All forces are specifically excluded. This feels com­pletely natural, but it definitely requires an explanation.
As discussed in the pure torque explanation, a pure torque does not have a point of application. It does not apply a net force, and it does not need a force support point to react against. It does require a torque reaction, but this does not apply in any particular point. This concept has surprisingly deep consequences, which can be put to very good use in simplifying both the exoskeleton itself and its use in practice. The details are discussed here.
The torque-only actuation is decoupled from all other movements, in particular from the translations of the shoul­der joint and of the torso. Not only are these not driven, but they are in fact left completely free floating.
This eliminates most of the disdavantages of conventional exoskeletons, in particular the lengthy and precise adjustment at the start of every session, but also the painful blocking of the natural synergies of the shoulder during abduction and other motions.
The shoulder abduction torque is driven along a horizontal axis coming from 45° back. This fits the human range of motion ( ROM ) quite closely, and since there is no translational forcing of the shoulder at all, the shoulder is free to lift and rotate around any point internal ( or even external ) to the shoulder.
This order of axes is far superior in range of motion to a vertical first axis from the world. This is often used in other exoskeletons because it provides natural gravity balancing ( or rather, indifference ) to the first axis, but this is not needed in an admittance controller, whcih only "sees" the weight suspended from the force sensor.
Designing a simple mechanism which can apply pure torques to the shoulder and elbow from a fixed base without constraining any of the other motions is not trivial. In the Armbot it is achieved by a relatively simple pantograph mechanism.
The mechanism decouples the torques from the translations by a 90° change in direction in some torque- and pushrods, via small bellcranks at the top of a swiveling column. The resulting exoskeleton does not fight, or even "see" any patient misalignments or any of the natural synergies such as the shoulder lifting during abduction. The video shows the idea. A separate page gives more details.
TODO Add the bit about no motors and the Galloway mechanism.
The torque-only idea was not the one that started the development. The original motivation was to use the skewed axes ideas from the ADL gimbal to eliminate the open circle segments, sometimes even big full ring bearings often seen near the human wrist interface, and even somemetimes even around the upper arm.
Eliminating these large bearings led to a mechanically so simple and lightweight robot arm.
, with no motors moving with the limbs, and with no cable drives or pulleys, only simple pushrods.
Several mechanical innovations were needed to make this ideal possible, as detailed below. But the major breakthrough came with the torque-only idea.
Upper arm abduction, flexion and internal-external rotation, as well as elbow flexion are powered without loading the shoulder joint and are absolutely painless. There is no setup or alignment time. Any patient can put their arm into the upper arm- and elbow brackets without any need for further adjustment.
Motors and force sensors are located in the base. The mechanism itself is very lightweight, consisting only of a few lightweight aluminium tubes. In addition, it is completely naturally balanced.
While at Moog, around 2009 I started the development of a novel arm exoskeleton. It is mechanically balanced. The mechanism gently floats when it is not powered. Bias forces ( or rather, torques ) are added to the mecha­nism by motors on the floor, with force sensors near the motor and admittance control to free up the motion completely.
By the use of a special elbow linkage ( modified Galloway ), ± 45° joint angles are transmitted over a series of joints with a total added angle of over 180°, all the way from the motors to the elbow and even to the wrist, by simple lightweight pushrods.
- skewed axes.
- abduction and eternal rotation mixed geometry.
- no vertical force on shoulder.
- torques only.
- free (unpowered) DOF's give shoulder joint 0.25 m free range in all directions.
- automatic balancing.
- additive angles via the Galloway bellcrank.
- low mass, low friction pushrods.
- all motors and force sensors at ground level.
- simple plain leadscrew drives.
- admittance control for extremely sensitive haptics.
- rubber disks take the place of small angle pivots.
- simple thrust bearings allow infinite pushrod roll.
- low cost Hudson Teknic ( ClearPath&hairps;) motors.
- one ( independent ) haptic channel per powered DOF.